Welcome
Welcome to Atomic Football, the home of what may be the only
truly objective college football rankings anywhere. While our
claim obviously leaves some wiggle room, you might still find it
somewhat bold. After all, all computer rankings are objective,
right? Not really. Let us explain.
The Problem
One only needs to wade into the shallows of football rankings
before the problem of "strength of schedule" is encountered. But
just how important is it? Every other football ranking system of
which we are aware and for which sufficient details are available
have "solved" this problem with what we call a "knob." Like the
volume knob on a radio, it must be set to some value. And like
the volume of your radio, it is a matter of personal preference.
So what? you might ask. Let's dig a little deeper.
So you have your ranking algorithm with your strength of
schedule knob, and you've tuned it to produce rankings you feel
are "reasonable." But is it possible that the final setting was
influenced by how well your favorite team faired? Perhaps a weak
team in a relatively strong conference - crank up the "strength
of schedule." Or maybe a strong team in a relatively weak
conference - turn down the "strength of schedule." Your rankings
are still only as unbiased as you are.
The Solution
Is there another way? What if you could find an algorithm that
knows how to tune itself. Well, after many years, we have finally
achieved this very thing. And, on top of that, we have shown how
to validate the tuning from statistics we can derive
independently of the algorithm - a self-consistency check, if you
will. If you're not into the math, then feel free to go straight
to the rankings page. Otherwise, please check out our paper.
Be aware that we don't do preseason rankings -- it just isn't
consistent with our obsession with objectivity. You will have to
wait until about the fourth or fifth week of the season before
the rankings will be published.
Our objective has been to develop a model-based approach to
the problem known to mathematicians as "ranking by pairwise
comparison." What makes football rather unique among sports, and
college football in particular, is the relatively small number of
games. Given also movement in recent years to de-emphasize
"margin of victory" and limit rankings to using only wins and
losses (a favorable trend, in our opinion), the problem has not
necessarily become easier. Using only wins/loss information, each
game is reduced to a single bit of information. Thus, for the
combined set of about 716 FBS, FCS, Division II, III, and NAIA
teams scheduled to play about 3720 games this year (2006), we
will have 3720 bits of data at the conclusion of the season. That
is the equivalent of 465 bytes of information -- about the same
amount of information as one verse of The Star-Spangled Banner.
From this tiny bit of data, we hope to accurately rank 716 teams.
Is it any wonder that this is such a controversial problem? To
learn more about the problem and our approach to solving it,
check out the Algorithm
page.
Rankings and Predictions
Our original rankings were a combination of two components --
a win-loss based ranking and a score-based ranking. Together they
are now known as our hybrid rankings and are meant to approximate how
typical fans might rank teams. They can be tracked on Ken Massey's
Ranking Comparison (FBS,
FCS)
under the symbol "ASH." The win-loss component of those
rankings has evolved over the years into what we originally called
our BCS-compliant rankings but we now designate as our win-loss rankings.
It is these rankings that are documented in great detail in the paper
you will find on our "Algorithm" page. These rankings appear on
Ken Massey's Ranking Comparison Page under the symbol "ABC."
More recently, our score-based rankings (which do not appear as
independent rankings on our website) have taken on a more predictive
flavor and are used as the basis for our score predictions. In our
first year (2007) on
Todd Beck's Prediction Tracker, we finished first in a field of
61 competitors for accuracy for the second half of the
season (see
College Football Ratings PT Awards). Last year (2008), we trailed
only the betting line in points accuracy for the season. We're not
yet bold enough to declare our predictions the best in the country, but
we're getting close.
Some Other Notes
Note that we have no affiliation with the BCS. Our win-loss
rankings are "BCS compliant" to the extent that they conform
to the best of our knowledge to constraints imposed by the BCS on
the various computer ranking systems that compose it
(e.g., using only wins and losses -- no scores).
They do NOT necessarily represent the BCS computer rankings
as they might appear were they generated for divisions other than the FBS.
Since many of the BCS Computer Ranking formulas are secret (or at
least insufficient details are available to reproduce them precisely),
BCS-equivalent rankings of the divisions beyond the FBS do not exist.
Also, be aware that while the constraints imposed by the BCS on their component
computer rankings work well for the FBS, they are not as
well-suited for the other divisions due to their poor "connectivity"
(i.e., more regional play) and will tend to produce less accurate
results until later in the season.
Jim Ashburn
11 March 2005
Last Updated 13 May 2009
|
